Warning : many lost ticks, time source seems to be instable

pc486 · n00b Joined: 28 Aug 2004 Posts: 1

I was experiencing this problem with a Tyan S2882 board (Thunder K8S Pro). I solved it by including in the kernel (instead of a module) support for the AMD-8111 contoler. No more lost ticks.

FYI, I'm using the gentoo-dev-sources 2.6.7-r14 something-or-rather.

andromexus · Posted: Wed Sep 01, 2004 1:06 pm Post subject: Desktop PC: No overclocking

Here, the same problem ocurred, it even sometimes stopped the machine (while emerging). But it was easily solved by disabling dynamic overclocking feature in the bios (my mobo is a MSI-Neo).

So better would be overclocking by software (who knows good examples for that?)

PrakashP · Posted: Wed Sep 01, 2004 1:32 pm Post subject:

No you need static overclocking - if you want to oc, no matter if it is software or hardware.

amd64gentoo · n00b Joined: 02 Sep 2004 Posts: 30

I noticed an above post that mentioned compiling support for the AMD 8111 into the kernel. I can't find any documentation that indicates that my mobo uses the amd 8111 so, I doubt that is the only source of this problem however, I found this interesting comment in the kernel source.

In the linux kernel source, file

/usr/src/linux/arch/x86_64/kernel/time.c

the comment on line 229 reads

/* AMD 8111 is a really bad time keeper and hits this regularly.
It probably was an attempt to avoid screwing up DST, but ignore
that for now. */

This does not give me a good feeling.

crazycat · Posted: Thu Sep 02, 2004 8:03 pm Post subject:

if someone lives in germany there is an article in latest c't about a problem of acer notebooks and linux with this hpet timer.

amd64gentoo · n00b Joined: 02 Sep 2004 Posts: 30

For me, none of the suggested "fixes" for this problem worked. Using the latest kernel source hides the problem but if I use the "repost_lost_ticks=100" boot parameter then they reappear. Is Gentoo running my processor much harder than it needs to run? Am I not getting the most out of my processor because I'm throwing away processor cycles?

Do any of the other linux distros have this problem? Mandrake 10, Suse 9.1 pro?

Here are some URLs discussing other people's linux "many lost ticks"

Subject: APIC problems on Via system
http://www.x86-64.org/lists/discuss/msg05593.html

"The new lost-tick detection code"
http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.61/2.5.61-mm1/broken-out/report-lost-ticks.patch

"Timer and APIC problems on x86_64 with 2.6.7"
http://www.ussg.iu.edu/hypermail/linux/kernel/0406.3/1073.html

amd64gentoo · n00b Joined: 02 Sep 2004 Posts: 30

I tried the report_lost_ticks=100 boot parameter in Mandrake 10.0 for amd64 and it does not report any lost ticks. What is different between Mandrake 10.0 and Gentoo?

mars-red · Posted: Wed Sep 15, 2004 12:50 pm Post subject: another victim

Hi all,

I just completed a stage 1 install on my new socket 939 amd64 (w/ K8T800 chipset mobo). I've been lurking the forums for a week to help get myself through the install, but this is my first post here.

I'm having the same lost ticks / instable clock messages. I am not running squat right now, just a bare console environment with a farked NIC driver. I have gathered the info from this thread and will do some experimenting tonight when I get home. I will post with my results. **fingers crossed**

-max
_________________
my wrench blog: http://www.digitaldownpour.com/blog

vyzivus · Posted: Tue Sep 21, 2004 10:21 am Post subject:

Well I upgraded to kernel 2.6.8-gentoo-r4 and some other messages have emerged:

JonSvenJonsson · n00b Joined: 24 Oct 2003 Posts: 56

Same problem here. I got an msi k8 neo. I solved my problem by adding a patch which lets powernow-k8 run on buggy bioses (before the patch i had a message that numpst must be 1 when inserting powernow-k8 module).

The patch can be found here:
http://www.muru.com/linux/amd64/patches/patch-2.6.4-powernow-k8-buggy-bios

However it wont apply cleanly against newer kernels. So you have to fix it, or wait until i have acces to my amd64-box, so i can make a diff oy my modifications for linux-2.6.9-rc2.

cu Jon

vyzivus · Posted: Sun Sep 26, 2004 2:10 pm Post subject:

Yo! I was playing with ACPI a little and this
is what I have found. Recently, I was having the 'many lost ticks' message
only once per minute, after I quit using GKrellM. I have searched for the source
of these messages and I found it: the Gnome battery meter. When I turned it
off the messages stopped. There is only one such message in my dmesg:

If you had the message then you may try

db9052 · n00b Joined: 11 Jun 2004 Posts: 5

odondor · n00b Joined: 29 Jun 2004 Posts: 4 Location: Belgium

I can confirm that it is battery related. After quitting gkrellm and the kde battery monitor the problem promptly disappeared. Good to have logging useable after 3 months.

bluz · n00b Joined: 18 Aug 2002 Posts: 61 Location: Canada

same here. I removed the gnome battery applet from my panel and the messages disappeared... weird.

UPDATE: Lost ticks returned... gnome applet not the issue, sorry I jumped the gun.

PC_Freak · n00b Joined: 29 Dec 2003 Posts: 63

I have had the lost ticks problem very recently. It became fixed when I was searching a way to get DMA enabled on the sata disk. I discovered that support for the SiS 5512 northbridge chip wasn't enabled in the kernel. After recompiling, DMA worked perfectly and the lost ticks problem was gone.

Even more, the kernel announced this

malloc · l33t Joined: 19 Sep 2003 Posts: 762

Well i'm having this problem with 2.6.9-ck3 and i think that (in my case) i suspect what's going on.
This warning only appears once during boot, right after my usb mouse is recognized (meaning after the load of ehci wich i have built-in into the kernel). This warning didn't appeared before and now that i think of it i've changed something between ck2 and ck3.
There's an USB option wich is something like poll time something, wich basically allows you to change the mouse poll timing from the default 100ms to something else. Now according to the docs my mx500 should support a poll interval of 2ms (50x lower than the default) wich is what i believe is causing the problem (after all it says "some driver may be hogging the interrupt resource").
I'm gonna try and re-compile my kernel without that option and see what comes out of it.
Anyway just an informative update.
_________________
--> Linux ### 2.6.11-ck2 #1 Sat Mar 12 20:21:30 WET 2005 i686 GNU/Linux <--

vyzivus · Posted: Sun Nov 14, 2004 7:43 pm Post subject:

PrakashP · Posted: Thu Nov 18, 2004 12:45 pm Post subject:

From Andi Kleen on lkml:

andersenep · Posted: Sun Jan 02, 2005 8:59 pm Post subject:

vyzivus

Interesting... I don't remember changing anything in kernel, just upgrading to 2.6.9-gentoo-r12, make oldconfig and the messages have disappeared. Could it mean that the problem got fixed in the kernel?
What do you mean by via82xx? Do you mean the IDE VIA82CXXX chipset support? I had it compiled not as a module but the 'lost ticks' messages were still present.
_________________
I thought what I'd do was, I'd pretend I was one of those deaf-mutes or should I?

scharkalvin · Posted: Tue Jan 04, 2005 3:37 pm Post subject:

redflo · n00b Joined: 17 Jan 2005 Posts: 1

did someone ever try the solution here:

http://community.linux.com/community/04/09/23/1624247.shtml?tid=11&tid=51&tid=95

They say, adding the kernel option "clock=pmtmr" solves the problem.

Al_Kane · n00b Joined: 19 Jan 2005 Posts: 13

DiscreteOnyxOrb · n00b Joined: 08 Apr 2004 Posts: 24

I have a k8v deluxe. I turned power management support and ACPI off in the kernel (dev-sources-2.6.7.r14). I have VIA82CXXX compiled in. I have always gotten this message, though changing the kernel parameters seems to change the frequency. I actually think vmware exacerbates it, but I can't be sure.

If the root of this problem really is ACPI, shouldn't turning off kernel support for it stop the symptom?

Anyone experiencing this on a 2.6.10 kernel?

CPU0
0: 528876506 XT-PIC timer
1: 34 XT-PIC i8042
2: 0 XT-PIC cascade
5: 3019411 XT-PIC libata, ehci_hcd, VIA8233, eth0
8: 0 XT-PIC rtc
10: 688815 XT-PIC libata, uhci_hcd, uhci_hcd
11: 107863345 XT-PIC uhci_hcd, uhci_hcd, nvidia
14: 936356 XT-PIC ide0
15: 2916691 XT-PIC ide1
NMI: 57205
LOC: 528779633
ERR: 250
MIS: 0